Modelling the transmission and spread of yellow fever in forest landscapes with different spatial configurations

Abstract

Yellow fever (YF) is a major public health issue in tropical and subtropical areas of Africa and South America. The disease is caused by the yellow fever virus (YFV), an RNA virus transmitted to humans and other animals through the bite of infected mosquitoes (Diptera: Culicidae). In Brazil and other South American countries, YFV is restricted to the sylvatic cycle, with periodic epizootic outbreaks affecting non-human primate (NHP) populations and preceding the emergence of human infections in areas close to forests. In recent epizootic-epidemic waves, the virus has expanded its range and spread across highly fragmented landscapes of the Brazilian Atlantic coast. Empirical evidence has suggested a possible relationship between highly fragmented areas, increased risk of disease in NHP and humans, and easier permeability of YFV through the landscape. Here, we present a hybrid compartmental and network-based model to simulate the transmission and spread of YFV in forest landscapes with different spatial configurations (forest cover and edge densities) and apply the model to test the hypothesis of faster virus percolation in highly fragmented landscapes. The model was parameterized and tested using the pattern- oriented modelling approach. Two different scenarios were simulated to test variations in model outputs, a first where the landscape has no influence on model parameters (default) and a second based on the hypothesis that edge density influences mosquito and dead-end host abundance and dispersal (landscape-dependent). The model was able to reproduce empirical patterns such as the percolation speed of the virus, which presented averages close to 1 km/day, and provided insights into the short persistence time of the virus in the landscape, which was approximately three months on average. When assessing the speed of virus percolation across landscapes, it was found that in the default scenario virus percolation tended to be faster in landscapes with greater forest cover and lower edge density, which contradicts empirical observations. Conversely, in the landscape- dependent scenario, virus percolation was faster in landscapes with high edge density and intermediate forest cover, supporting empirical observations that highly fragmented landscapes favour YFV spread. The proposed model can contribute to the understanding of the dynamics of YFV spread in forested areas, with the potential to be used as an additional tool to support prevention and control measures. The potential applications of the model for YFV and other mosquito-borne diseases are discussed.